(169) Quantification of information gained by linking claims data to an EHR cohort of US metastatic breast cancer (mBC) patients

Background: Linking patients’ claims data to their EHR can improve ascertainment of confounders at baseline and events during follow-up, at a cost of lower sample size. Quantifying information gained and differences in patient characteristics from EHR to EHR-claims cohorts informs study design.

Objectives: In mBC patients, compare EHR vs EHR-claims cohorts regarding
1. Lookback and follow-up time
2. Unique diagnoses
3. Baseline characteristics
4. Rates of adverse events (AEs)

Methods: A retrospective cohort study used ConcertAI Patient360 EHR linked to multiple closed claims sources (including Medicare Advantage). Patients with incident mBC were included in an EHR cohort, and a subcohort had ≥ 90 days of claims coverage (01 October 2015 to 30 April 2023). Index date was the first mBC diagnosis; follow-up ended at death, end of data availability, last activity (EHR), or end of enrollment (EHR-claims). For each patient, numbers of unique diagnoses overall and per month were calculated. Patient-time during lookback and follow-up, baseline demographics and clinical characteristics, including Charlson comorbidity index (CCI, excluding cancer), and for 14 AEs rates per 100 person-month (p-m) and incidence proportions (IP) at 180 days were calculated. Because many patients change coverage at age 65, analyses were age-stratified.

Results: For the EHR cohort N = 6289; 1438 (23%) were in the EHR-claims subcohort. A greater % were aged ≥65 years in the EHR cohort (30%) than in the EHR-claims cohort (17%). ¬¬
Lookback and follow-up periods were longer in EHR-claims than in the EHR cohort. Age < 65: median 868 vs 691 days lookback; 976 vs 734 days follow-up. Age ≥65: median 902 vs 462 days lookback; median 880 vs 574 days follow-up.
In both age groups, the median unique diagnoses per person and per p-m were higher in the EHR-claims than in EHR. For example, in < 65, the median per person was 36 in EHR-claims and 7 in EHR. Of patients < 65, 80% of patients in EHR and 66% in EHR-claims had CCI of 0.
For most AEs, rates were higher in both age groups in EHR-claims than in EHR (in < 65, from 4.8/p-m higher for neuropathy to 0.2/p-m lower for death; in ≥65, from 3.5/p-m higher for anemia to 0.7/p-m lower for death). IPs were also higher in EHR-claims for most AEs.

Conclusions: EHR-claims provide more diagnoses and observation time, at the cost of a reduction in sample size and underrepresentation of patients ≥65. Age-stratified or weighted analyses in EHR-claims data are recommended. These results aid in interpretation of differences between EHR and claims results where shifts in age, completeness of diagnosis history, and duration of observation combine to influence results.